CN113673389B

CN113673389B - Painting illumination visual evaluation method related to spectrum power distribution of light source

Info

Publication number: CN113673389B
Application number: CN202110907412.3A
Authority: CN
Inventors: 党睿; 康亚芳
Original assignee: Tianjin University
Current assignee: Tianjin University
Priority date: 2021-08-09
Filing date: 2021-08-09
Publication date: 2023-07-18
Anticipated expiration: 2041-08-09
Also published as: CN113673389A

Abstract

The invention relates to the technical field of the intersection of spectral composition and illumination quality evaluation of LEDs, which aims to establish the direct connection between a light source SPD and a subjective evaluation result and determine a spectral specific wave band with obvious influence on the illumination quality evaluation result; providing a reference for the design of a multi-channel LED light source in a museum display environment; providing a foundation for revising the lighting standards of the museum; for the drawing illumination visual evaluation method related to the spectrum power distribution of the light source, aiming at a museum drawing cultural relic sample, firstly, establishing a direct relation between spectrum power distribution SPD and a subjective evaluation result through a visual evaluation experiment, and determining a spectrum specific wave band with obvious influence on an illumination quality evaluation result; and then, establishing a museum drawing illumination quality evaluation model based on SPD through data analysis and mining, so as to realize drawing illumination visual evaluation. The invention is mainly applied to the occasion of evaluating the illumination quality.

Description

Painting illumination visual evaluation method related to spectrum power distribution of light source

Technical Field

The invention relates to the technical field of intersection of spectral composition and illumination quality evaluation of LEDs, in particular to a method for acquiring visual evaluation effects of museum painting illumination related to spectral power distribution of a light source.

Background

1. Background

The total number of Chinese museums in 2018 reaches 5136, wherein the number of Chinese traditional drawing collections is 70 tens of thousands of pieces, and the storage is huge. Meanwhile, the Chinese painting has rich picture content, various colors and very high cultural and artistic value, and is not only a precious component part of Chinese history, but also an important element of world cultural heritage. The investigation of corrosion loss of the cultural relics collected in the whole country shows that the cultural relics drawn in the country have different degrees of damage, which is mainly related to the preservation environment of the cultural relics in the house, including temperature, humidity, illumination, air pollutants, microorganisms and the like. The museum side can adjust the temperature and humidity and the air pollutants to the optimal state required by the preservation of the cultural relics in the collection through installing proper equipment. However, as an indispensable factor for the exhibition and research of the cultural relics in the collection, any illumination radiation may cause illumination damage to the cultural relics in the collection. In order to reduce the damage caused by illumination, the related specifications prescribe the illumination quantity index (such as illumination, exposure and the like) of the light source in the house, and meanwhile, some students also develop researches on illumination damage of painting exhibits. The above specification and study on illumination damage of cultural relics is helpful to make a proper and accurate Wen Bao strategy, but at the same time, the improvement of illumination quality of cultural relics is also widely accepted. Therefore, while enhancing the protection of cultural relics, related researches for improving the illumination quality of the cultural relics are required to be carried out.

In subjective evaluation research on illumination quality, students at home and abroad determine ornamental evaluation indexes from two factors of light source color appearance and light source quality. Wherein the light source color appearance is typically expressed in terms of correlated color temperature. However, considering that the light source has metamerism, even the spectrum of the same correlated color temperature has a difference, so that the correlated color temperature is difficult to comprehensively characterize the spectrum characteristics. Common indicators of light source quality include Color Rendering Index (CRI), color Gamut Area Index (GAI), color quality value (CQS), memory Color Rendering Index (MCRI), color rendering index CRI-CAM02-UCS, IES TM-30 (a light source color quality evaluation method), and the like. Due to the range of applicability differences and the respective limitations between these indices, it is difficult for a single index to fully characterize the quality of a light source. In contrast, the spectral power distribution (Spectral Power Distribution, SPD) is a fundamental property of the spectral composition of a light source, which is decisive for its color appearance and quality. As the common light source for the illumination of the cultural relics in the collection of the cultural relics, the spectrum of LED (Light Emitting Diode) is very flexible, and the spectrum of LEDs with different manufacturing principles, different manufacturers and different models has huge differences, so that the illumination quality of the cultural relics can be obviously influenced. Meanwhile, the LED is used as a multi-channel light source, and the light source SPD meeting the requirements can be obtained by adjusting the light emitting intensity of each channel, so that the characteristics provide possibility for designing the light source with illumination protection and illumination quality. Therefore, as the application of LEDs in museum illumination is becoming wider, it is necessary to explore the law between the light source SPD and the illumination quality of the cultural relics in the collection, and to determine and design the optimal light source SPD by taking the improvement of the illumination quality of the cultural relics as a guide.

In summary, in the context of the current illumination quality urgent need to be improved, the present invention provides a method for evaluating the illumination quality of museum painting related to the light source SPD. The invention can help the museum to improve the painting illumination quality and can also provide a certain guidance for revising the illumination standard of the museum.

2. The current closest prior patents to this application are as follows:

(1) A bronze ware spread Chen Zhaoming light quality evaluation method and system (patent number: CN 201911006731.6). The invention discloses a bronze ware spreading Chen Zhaoming light quality evaluation method and system. For the light source to be evaluated, the patent obtains corresponding estimated values according to the chromaticity information and the illumination quality estimation model, and realizes characterization of the bronze ware spread Chen Guangyuan illumination quality. As mentioned earlier, SPD as the fundamental characteristic of a light source determines the color appearance of the light source, the quality of light color, and its chromaticity coordinates, but the patent only evaluates the illumination quality of the light source to be measured from the chromaticity information of the light source to be measured. Meanwhile, the patent is not directed to evaluating the illumination quality of museum drawings, and cannot provide a beneficial reference for the illumination quality of museum drawings at present.

(2) A multi-level illumination color temperature preference prediction method and system (patent number: CN 202011430644.6) for national painting Chen Zhaoming. The invention discloses a multi-stage color temperature illumination preference prediction method and system for national painting exhibition Chen Zhaoming. The patent mainly combines an illumination preference estimation model to evaluate the illumination quality of a light source on the basis of screening the surface illumination (between 100lux and 1000 lux) of the traditional Chinese painting to be displayed and the color temperature (between 2500K and 5500K) of the light source to be evaluated, and still starts from two indexes of relevant color temperature and illumination, does not study from the fundamental characteristic angle of the light source SPD, and cannot establish a specific influence relation between the spectrum and visual evaluation according to the illumination quality.

(3) The invention discloses a method and a system for evaluating the quality of spanned light Chen Zhaoming based on illuminance and correlated color temperature (patent number: CN 202010334650.5). The invention discloses a method and a system for evaluating the quality of spanned light Chen Zhaoming based on illuminance and correlated color temperature, which realize characterization of the quality of spanned light source based on the whiteness and color preference of the light source, so as to judge whether the illuminance and the color temperature of the light source to be measured are in a set illuminance and color temperature range. The invention only starts from illumination and correlated color temperature to evaluate the light quality of the display Chen Zhaoming, has certain limitation, and does not take museum drawing illumination as a research object.

3. The studies closest to the present application are as follows:

(1) Spectrum optimization design and color quality evaluation method research [ D ] based on multichannel LED light source]Zhang Fuzheng doctor of Zhejiang university sets four correlated color temperature levels under 450lx illuminance, and each color temperature uses a multi-channel LED light source to obtain metamerism spectrum defining values of Rf (color fidelity index) and Rg (color gamut area index), which are used as experimental conditions. A multichannel LED light source is assembled on the ceiling of a room, articles with various colors and Macbeth color cards are placed, and an observer scores the favorability, naturalness and vividness of the articles under each working condition. Through statistical analysis, the correlated color temperature can be obtained without significantly influencing the color quality evaluation, and the characterization effect of the existing color quality evaluation index on the light source is limited, thereby establishing the R-based color filter _f 、R _g R is R _cs,h1 Color quality prediction model of three parameter combinations.

The study selects three indexes in light color quality only to evaluate the illumination quality of the light source, and has certain limitation.

(2) Light environment evaluation and related LED spectrum research [ D ]. The university of Chinese patent application Zhou Mengxin discloses that by optimizing light source spectrum, blue light hazard is reduced or color gamut area is increased under the condition that color temperature of the light source is kept unchanged, and visual experiments of subjective color vividness and preference under illumination and spectral matching after the color gamut area change of the light source are carried out, subjective evaluation experiments are carried out on 6 light sources with different spectral structures with the same color point and the same color temperature and the same color gamut area, and statistical analysis is carried out on subjective evaluation results of influence of the light source.

The study proves that the spectrum structure and the color of the color block have significant influence on subjective evaluation, but a mathematical surface model between the spectrum and the color block and the subjective evaluation is not established on the basis of the subjective evaluation, and the specific relation between the spectrum structure and the color block and the subjective evaluation cannot be quantified.

(3) The research of the comprehensive evaluation method of the illumination quality of the commercial building [ D ]. The Bo Yanhui master of Tianjin university carries out the field investigation on two large commercial complexes of Beijing, selects the physical quantities of illumination, general color rendering index, correlated color temperature and the like to carry out the actual measurement, and carries out the subjective evaluation of the visual comfort, thus finally obtaining the influence rule of the color temperature on the visual acuity, the influence rule of the illumination on the color vision and the influence rule of the illumination parameter on the illumination quality.

The study still evaluates the illumination quality from the illumination and color temperature perspective, fails to realize a study of the law of influence between spectrum and illumination quality, and the study is not directed to evaluating the museum drawing illumination quality, failing to provide a beneficial reference for the current museum drawing illumination quality.

4. The national standard currently relevant to the present application is the "museum lighting design Specification" (GB/T23863-2009), in which the relevant regulations for the light source selection of Chinese traditional relic lighting are as follows:

(1) The 4.3.1 th light source with the color temperature smaller than 3300K is used as the illumination light source.

(2) In the 4.3.2 th place with high color discrimination requirement such as display drawing, color fabric, multicolor exhibit, etc., a light source with a general color rendering index (Ra) not lower than 90 should be used as the illumination light source.

(3) The 6.0.1 light source should reduce ultraviolet radiation and infrared radiation so that the relative ultraviolet content of the light source is less than 5 mu W/1m.

However, this standard has the following problems:

current standards began to execute in 2009, and no new revisions exist. Until now, LEDs have not been used in museum lighting, and thus the current standard is related index established for conventional light sources such as metal halogen lamps, tungsten halogen lamps, fluorescent lamps, and the like. However, LEDs have been rapidly developed in recent years as a new generation light source, and are beginning to be used in a large number of museum illuminations. The LED light source can meet the requirements of three indexes of no ultraviolet and infrared, color temperature lower than 3300K and general color rendering index higher than 90 in the standard, but the SPDs of different light sources have huge differences, and the corresponding illumination quality is different. Meanwhile, the spectrum of the LED is very flexible, and the spectrum of LEDs of different manufacturing principles, different manufacturers and different models has huge difference, so that the illumination quality of cultural relics can be obviously influenced. However, the regulations related to light source selection in the standard do not relate to the related requirements of the SPD of the LED, which makes the current standard unable to directly select the illumination light source most suitable for the exhibited article from the spectrum angle, thereby improving the illumination quality.

In summary, under the background that the lighting quality of the painting in the current museum needs to be improved, the existing research focuses on the development of lighting damage and protection of the painting exhibit. Meanwhile, in the existing subjective evaluation study on illumination quality, more indexes with low correlation between the appearance of light source color and the quality of light source are used for evaluation. And the related specifications cannot be revised with the update of the LED light source technology. Therefore, for the current research and standard, the illumination light source suitable for painting cultural relics cannot be selected from the fundamental characteristic of the light source, namely spectrum composition, with the aim of improving illumination quality without causing additional illumination damage.

Disclosure of Invention

To overcome the deficiencies of the prior art, the present invention aims to:

1. and establishing direct connection between the light source SPD and the subjective evaluation result, and determining a spectrum specific wave band with obvious influence on the illumination quality evaluation result.

2. An evaluation model of painting illumination quality of a museum based on SPD is provided.

3. Providing a reference for the design of the multi-channel LED light source in the exhibition environment of the museum.

4. Providing a foundation for revising the lighting standards of the museum.

For museum painting cultural relics, a visual evaluation experiment is firstly carried out to establish a direct relation between spectrum power distribution SPD and subjective evaluation results, and a spectrum specific wave band with obvious influence on illumination quality evaluation results is determined; and then, establishing a museum drawing illumination quality evaluation model based on SPD through data analysis and mining, so as to realize drawing illumination visual evaluation.

The method comprises the following specific steps:

1. using LEDCube as illumination source, operating by computer software LEDCNavigator V6.3.7 0606;

2. the ten narrowband spectra were used for spectral construction and iteration to obtain the final 40-item target spectrum, the parameters of the ten narrowband spectra are as follows:

the ten kinds of narrow-band spectrums are endowed with different irradiance values and are overlapped to form spectrums with different shapes, and the overlapping principle is as follows:

wherein n=1, 2,3, …,10, i=1, 2,3, …,6 irradiance values of each narrowband light source, S (λ) is spectral power distribution obtained by superposition, and Sn (λ) is irradiance 1W/m of each narrowband light source ² Spectral power distribution at time, A _i The irradiance of each narrow-band spectrum is valued, and the initial irradiance is set to be 0W/m ² The end irradiance was set to 10W/m ² At 2W/m ² Obtaining 6 irradiance values for the incremental step size of irradiance;

screening modes meeting target spectrums are screened:

screening step one: the range of the correlated color temperature is 2650K less than or equal to CCT less than or equal to 4150K, ra is more than or equal to 90, R9 is more than or equal to 0, and Duv is less than or equal to 0.0054;

screening step two: taking the irradiance value of the spectrum as a horizontal axis and the correlated color temperature corresponding to the irradiance value as a vertical axis to obtain the correlated color temperature distribution condition of a spectrum set with the same irradiance value, and taking the most uniform part of the correlated color temperature distribution of the spectrum with equal irradiance;

Screening step three: dividing the correlated color temperature range into a plurality of proper intervals, selecting 6-7 spectrums for each interval to obtain a final spectrum, and taking all irradiance values of each narrow-band light source for a target spectrum:

3. acquisition of any narrowband spectral area:

selecting 4 light emitting percentages of 25%, 50%, 75% and 100% for each narrow-band spectrum, measuring the actual irradiance, fitting the function of the irradiance and the light emitting percentages, wherein the horizontal axis represents the light emitting percentage, the vertical axis represents the ratio of the actual irradiance to the irradiance when the light emitting percentage is 100%, and calculating the area of any narrow-band spectrum under any working condition through the function;

4. establishing two-dimensional graphs of areas of narrowband spectrums with different peak wavelengths and evaluation scores according to experimental data, further determining specific spectrum bands influencing the evaluation through correlation analysis, establishing two-dimensional graphs of different colors and evaluation scores, and further determining specific colors influencing the evaluation through correlation analysis;

5. data processing and data mining, wherein the number of variables is reduced to three dimensions through correlation analysis, and the normalization processing of the independent variables is realized through using a function MinMaxScaler in a support vector machine (support vector machine, SVM); based on the conversion of the dependent variable into the label variable, the model establishment is further converted into three types of problems, and the comfort evaluation model, the definition evaluation model and the vividness evaluation model are established through data division and model training.

Establishing a comfort degree evaluation model, a definition evaluation model and a vividness evaluation model through data division and model training, wherein the method comprises the following specific steps of:

(1) Data preprocessing

a converting dependent variables into tag variables

For definition scoring, dividing the definition scoring mean value under each working condition into the following three sections: -4 is less than or equal to the definition score mean value < -1, -1 is less than or equal to the definition score mean value < 3,3 is less than or equal to the definition score mean value less than or equal to 4, the three intervals represent "poor", "good" and "excellent", and dependent variables in the three intervals are respectively marked as "1", "2" and "3" and are used as tag variables; comfort level score, vividness score and so on, are also classified into three grades of "poor", "good" and "excellent";

b normalizing the independent variable

Mapping each dimension of data to a specific interval using a self-contained function MinMaxScale in the SVM, wherein the default interval is [0,1];

(2) Model building process

a dividing the data into training and testing sets.

Setting a training set: test set = 4:1, randomly partitioning the data;

b parameter selection

Based on an evaluation model of a support vector machine, a Gaussian kernel function RBF is selected, and the kernel function is in the form of:

K(x _i ,x _j )＝exp(-γ‖x _i -x _j ‖ ² )

wherein (x) _i ,x _j ) For training set data, x _i Is the characteristic vector of the data, x _j For the labels, gamma is a parameter which needs to be set manually, before model training, penalty factors C and gamma values which are real numbers larger than 0 are required to be specified, and K-fold cross validation and a grid-based method are used for adjusting the parameters to obtain values of C and gamma;

model c training

For the data of definition evaluation, the total number of characteristic variables is 3, namely the sum of the area value of a narrow-band spectrum with the peak wavelength of 445nm, the area value of a narrow-band spectrum with the peak wavelength of 505nm, 525nm and 555nm, and the sum of the area value of a narrow-band spectrum with the peak wavelength of 595nm and 610nm, the target variable is the definition evaluation level obtained through processing, the parameters C and gamma are the values obtained in the step b, the characteristic variable and the target variable of a training set are input, and a definition evaluation model based on a support vector machine can be obtained through training; for comfort evaluation data, the number of characteristic variables is 2, namely the sum of the area value of a narrow-band spectrum with the peak wavelength of 445nm and the area value of a narrow-band spectrum with the peak wavelengths of 505nm, 525nm and 555nm, the target variables are comfort evaluation levels obtained through processing, the number of characteristic variables in the vividness evaluation data is 1, namely the sum of the area values of the narrow-band spectrum with the peak wavelengths of 595nm and 610nm, the target variables are vividness evaluation levels obtained through processing, wherein the obtaining mode of setting the values of parameters C and gamma is the same as that of step b, the characteristic variables and the target variables of a training set are input, and a comfort evaluation model and a vividness evaluation model based on a support vector machine can be obtained through training;

(3) Mathematical model

Inputting characteristic variables of test set data into a trained model, calculating a predicted value, calculating the difference between the predicted value and a target variable of the test set through a function metrics.

And the painting illumination quality evaluation model based on the support vector machine established through the process predicts the overall experience when the painting is watched under any exhibition Chen Guangyuan.

The invention has the characteristics and beneficial effects that:

(1) The acquisition method of the spectrum versus color reduction calculation model provided by the invention has the advantages that the spectrum specific wave band with obvious influence on the illumination quality evaluation result is defined, the gap of the museum painting illumination quality evaluation from the aspect of spectrum power distribution in the past is filled, the spectrum power distribution is used as an experimental variable, the quantitative relation between the spectrum and the visual evaluation is directly established, the information loss during conversion from two-dimensional spectrum data to one-dimensional indexes is avoided, and the spectrum profile meeting the target is further directly obtained.

(2) The invention establishes a museum painting illumination quality evaluation model based on spectral power distribution on the basis of a spectrum specific wave band which has obvious influence on an illumination quality evaluation result, and the illumination quality evaluation grade can be obtained by inputting data of any given spectral power distribution into the model.

(3) The invention can provide a basis for revising the lighting standard of the museum.

In conclusion, the lighting quality of museum drawing can be greatly improved, and further, the viewing experience of museum drawing is improved, so that the method has great social benefits.

Description of the drawings:

fig. 1 is a technical scheme flow chart.

Fig. 2 is a schematic diagram and a live-action photograph of an experimental independent showcase.

Fig. 3 is a spectral power distribution diagram of 14 single-color LEDs at an output ratio of 100%.

Fig. 4 is a diagram of a LEDNavigator software interface.

Fig. 5 is a spectral power distribution diagram of ten narrow-band LEDs at the same irradiance.

Fig. 6 is a correlated color temperature and irradiance distribution of 9477 spectra.

FIG. 7 is a spectral power distribution diagram of a target operating condition.

Fig. 8 is a view of the hairpin flower beautiful women (left) and a view of the cottage mountain (right).

Fig. 9 is a graph of actual irradiance versus percentage of light output (for example, a narrowband spectrum having a peak wavelength of 635 nm.

Fig. 10 is a graph showing the influence of the narrowband spectral areas of different peak wavelengths on the comfort evaluation result.

(a)425nm (b)445nm

(c)475nm (d)505nm

(e)525nm (f)555nm

(g)595nm (h)610nm

(i)635nm (j)660nm

(k)670nm (l)705nm。

Fig. 11 is a graph showing the influence of the sum of narrowband spectral areas having peak wavelengths in the same color region on the comfort evaluation result.

(a) Purple (b) blue

(c) Cyan (d) Green

(e) Orange (f) red.

Fig. 12 is a graph showing the influence of the narrowband spectral areas of different peak wavelengths on the sharpness evaluation result.

(a)425nm (b)445nm

(c)475nm (d)505nm

(e)525nm (f)555nm

(g)595nm (h)610nm

(i)635nm (j)660nm

(k)670nm (l)705nm。

Fig. 13 is a graph showing the influence of the sum of narrowband spectral areas having peak wavelengths in the same color region on the sharpness evaluation result.

(a) Purple (b) blue

(c) Cyan (d) Green

(e) Orange (f) red.

Fig. 14 is a graph showing the influence of the narrowband spectral areas of different peak wavelengths on the vividness evaluation results.

(a)425nm (b)445nm

(c)475nm (d)505nm

(e)525nm (f)555nm

(g)595nm (h)610nm

(i)635nm (j)660nm

(k)670nm (l)705nm。

Fig. 15 is a graph showing the influence of the sum of narrowband spectral areas having peak wavelengths in the same color region on the vividness evaluation result.

(a) Purple (b) blue

(c) Cyan (d) Green

(e) Orange (f) red.

Fig. 16 is an influence law chart of influence of the association factor on the comfort evaluation.

FIG. 17 shows the correlation between the influence of the sum of the area values of narrow-band spectra having wavelengths of 445nm and 595nm and the area value of the narrow-band spectrum having a peak wavelength in the green region on the comfort evaluation.

(a) The peak wavelength is 445nm and the correlation of the area value of the narrow band spectrum in the green region.

(b) Correlation of area values of narrowband spectra having peak wavelengths of 445nm and 595 nm.

(c) Correlation of peak wavelength in green region and area value of 595nm narrow band spectrum.

Fig. 18 is a label distribution of experimental data.

Fig. 19 is a ROC curve (taking a definition evaluation model as an example).

Fig. 20 is a process diagram of applying a museum painting lighting quality assessment model.

Detailed Description

The acquisition method for the illumination quality evaluation model meeting the illumination visual requirements of the museum drawing has a large reference value for the design of the multi-channel LED light source in the exhibition environment of the museum, and has great significance for improving the painting and ornamental experience of the museum.

The technical scheme flow is shown in figure 1.

1. Experimental environment

(1) In order to truly simulate the painting and lighting environment of the museum, a real-scene museum exhibition hall is built in the physical environment of the Tianjin university building and the key laboratory of ecological technology in Tianjin city, the exhibition hall length and the exhibition width are 9m, and the height is 4.2m. The white emulsion paint is brushed on the surface of the inner wall, the reflectivity is 0.8, and the reflectivity of the cement floor is 0.2. To more truly restore the showroom, three types of 5 wall-mounted showcases and 5 independent showcases are provided, and imitation pictorial representations and ceramic products are displayed in the showcases.

(2) One of the column independent showcases was selected as the device for the evaluation experiment, and its size was 600×600×220mmj. The height of the base is 800mm, gray light-absorbing cloth is paved on the base, and painting works are placed on the base, so that influence of reflection on the surface of the base on visual evaluation results is reduced. And a key illumination light source is arranged above the showcase, and the distance from the bottom of the light source to the surface of the painting work is 1100mm. The schematic diagram and the live-action photo of the independent showcase are shown in figure 2

2. Experimental light source

(1) The light source for the accent illumination is a spectrally tunable illumination device LEDCube with dimensions 300mm by 210mm (length by width by height). LED cube consists of 14 LEDs of different peak wavelengths, and the spectral power distribution of each single-color LED is shown in fig. 3.

(2) The computer software LEDNAPHEToRV 6.3.7 0606 is used for operation, the LEDCube can be controlled and regulated to reproduce any input spectral power distribution, or the luminous intensity of any channel is regulated for light source design, and the brightness can be regulated on the premise of ensuring the same light quality, and the operation interface is shown in figure 4.

3. Experimental conditions

The experimental variable is the spectral power distribution and the spectrum used as the experimental regime is determined by the following two steps.

(1) And determining a target spectrum, namely a spectrum working condition in an ideal state. Finally, 40 target spectrums are obtained in total, and the specific operation method is as follows:

Ten narrowband spectra were used for spectral construction and iteration, and table 1 lists the parameters of the ten narrowband spectra, the respective spectral power distributions of which are shown in fig. 5.

Table 1 parameters of ten narrowband light sources

Any spectral power distribution can be obtained by superposition of narrow-band spectra of a plurality of different peak wavelengths according to the superposition of the spectra. Therefore, ten kinds of narrow-band spectrums are endowed with different irradiance values and are overlapped to form spectrums with different shapes, and the overlapping principle is shown in a formula (1).

Where n=1, 2,3, …,10, represents 10 narrowband light sources, i=1, 2,3, …,6, represents 6 irradiance values for each narrowband light source. S (lambda) is the spectrum work obtained after superpositionA rate distribution, sn (lambda) is the irradiance of 1W/m for each narrow-band light source ² Spectral power distribution at that time. A is that _i The irradiance of each narrow-band spectrum is valued, and the obtaining method is as follows: the initial irradiance of each narrow-band light source in the iterative process is set to be 0W/m ² The end irradiance was set to 10W/m ² At 2W/m ² For increasing irradiance step sizes, 6 irradiance values were obtained in total. All spectrum compositions obtained by superposition of ten narrowband light sources are exhausted to obtain 6 ¹⁰ A spectrum.

In a museum display environment, the light source selection is required to meet not only the requirements of cultural relics protection, but also the visual requirements of ornamental cultural relics. Therefore, the value range of correlated color temperature is set to be 2650K less than or equal to CCT less than or equal to 4150K, ra is more than or equal to 90, R9 is more than or equal to 0, duv is less than or equal to 0.0054 as a screening condition. Pair 6 ¹⁰ And (3) carrying out iterative calculation on the spectrums, and removing spectrums which do not meet the screening conditions, so as to finally obtain 9477 spectrums. And taking irradiance values of the spectrums as horizontal axes and corresponding correlated color temperatures as vertical axes, the correlated color temperature distribution condition of the spectrum set with the same irradiance value can be obtained, as shown in fig. 6.

As can be seen from the graph, when the irradiance is 40W/square meter, the correlated color temperature distribution of the spectrum of the equal irradiance is most uniform; thus, the target conditions were obtained by screening from a collection of all spectra (737 total) with irradiance of 40W/square meter. The spectrum is divided into 6 intervals according to the correlated color temperature range, 6-7 spectrums are selected for each interval, 40 spectrums are total, and the spectral power distribution in each color temperature interval is shown in table 2. It can be seen that the spectra in each color temperature interval are different, the positions and intensities of the peak wavelengths are different, and the 40 spectra can take all irradiance values of each narrow-band light source (see table 3), so the spectra are representative.

TABLE 2 spectral power distribution map selected for each color temperature interval

Table 3 irradiance values for ten narrowband light sources comprising a spectrum having an irradiance of 40W/square meter

And normalizing the illuminance values of the 40 spectrums obtained by screening within the range of 50-100lx, and taking the normalized illuminance values as a target experimental condition, wherein the obtained target spectrums are shown in figure 7.

(2) The target spectrum is reproduced by using the LEDCube, and the obtained reproduced spectrum has errors with the target spectrum due to the characteristics and the accuracy limitation of the equipment, so that the reproduced spectrum is taken as a real experimental working condition. The specific operation method is as follows:

to reproduce the target spectrum obtained in the first step, it is fitted using a LEDCube. Under the "light distribution" function menu, "SPD mode-Shape" is selected, i.e. matching is performed according to the inputted spectral Shape. And importing the power distribution of the spectrum to be fitted, and setting the illumination value to obtain the spectrum with similar shape. Because the peak wavelength of each single-color LED in the LED cube is not completely consistent with that of a narrow-band light source used in the earlier stage research, and the partially reproduced spectrum is different from the theoretical spectrum, the influence of the actual spectrum obtained by fitting is taken as a variable in the data analysis process, and the influence of the actual spectrum on the subjective evaluation result is researched. The color temperature range of the actual working condition is 2750-4100K, and the actual spectral power distribution obtained by fitting in each color temperature interval is shown in Table 3.

4. Experimental protocol

The traditional Chinese painting has various classification standards and can be roughly divided into two types according to the expression technique: painting and sketching. Therefore, two different techniques and drawing of the subject materials are selected as representatives for evaluation, namely a hairpin flower beautiful women drawing and a cottage mountain drawing. The painting is used as cyan ink, and the main pigment types used include ocher, lime, graphite and the like. The painting used in the experiment is a customized imitation, and is subjected to special process treatment, the color appearance of the painting is similar to that of a genuine product, and the painting can be subjected to visual evaluation research according to the color appearance of the painting, which is shown in fig. 8.

A total of 12 subjects were enrolled, 6 men and 6 women, aged between 23-27 years, with an average age of 24.7 years. All the tested passed the Ishihara color vision test to correct vision. Before the start of the experiment, the meaning of the experimental procedure and each evaluation index was solved for the test.

The evaluation content includes an atmosphere perception of the overall lighting environment, characterized by comfort, and an appearance of the drawing, characterized by sharpness and vividness. The evaluation index includes three pairs of anti-ambiguous words: comfort-discomfort, clarity-blur, vivid color-monotonous color. Comfort refers to the overall coordination, clear refers to whether the details and textures of the exhibit can be clearly displayed, and vivid color refers to the overall color feel. The meaning of each score and its corresponding meaning is shown in table 4.

Table 4 classification rating scale

In the experimental process, the recorder and the tested person both wear the black experimental clothes to reduce the influence of the reflection of the clothes on the evaluation result. The tested person stands at the position 300mm in front of the showcase, views the painting, and evaluates the whole visual effect. The time for each trial to participate in the experiment was sixty minutes.

5. Spectral data processing

The nature of the fitted spectrum consists of a superposition of narrow-band spectra of 14 different peak wavelengths, so that the spectral variation can be described by the area values of the 14 narrow-band spectra. However, due to the characteristics of the device itself, the actual light output brightness is not linearly related to the light output percentage shown at the interface. Thus, for each narrowband spectrum, 4 percent light out of 25%, 50%, 75%, 100% were chosen, the actual irradiance was measured, and the irradiance was fitted as a function of percent light out, see fig. 9. Wherein the horizontal axis represents the percentage of light and the vertical axis represents the ratio of the actual irradiance to the irradiance at 100% of the percentage of light. By this function, the area of any narrowband spectrum under any operating condition can be calculated.

In the process of spectrum reproduction, the light emission percentages of the narrow-band spectrums with the peak wavelengths of 405nm and 540nm are 0, namely, the narrow-band spectrums are not used in the light distribution process, so that the final variable is the area value of the narrow-band spectrums with 12 different peak wavelengths. Table 5 shows some of the operating conditions, and it can be seen that the area values of the narrowband spectrum for each peak wavelength may be the same or different for different operating conditions. This is affected by the fitting process of the LEDCube to the target spectrum, and for each narrowband spectrum of peak wavelength, different conditions may be fitted using the same area value, which has no effect on experimental data analysis. During this conversion, the data dimension of the spectral power distribution is not lost, and it can still be represented as a two-dimensional matrix.

TABLE 5 area values for each narrowband spectrum under partial operating conditions

6. And analyzing the rule of influence of area values of narrowband spectrums with different peak wavelengths on three degrees.

(1) Data preprocessing

Outliers that are present in the experiment and affect the results of subjective evaluations are removed using the quartile range method. To represent the inter-observer differences, the standard deviation of the scores of the three rating scales under each condition was calculated. The results showed that the standard deviation ranged from 0.5 to 2.4. Table 6 lists the standard deviation averages of the scores of the three evaluation scales. It can be seen that the results of the three evaluation scales have better consistency between different subjects. Thus, the average score of the observer was used for the next analysis.

Table 6 standard deviation of the score of the rating scale

Two drawings with different techniques are selected in the experiment, namely a painting drawing and a sketching drawing. Correlations between painting types and rating scale scores were analyzed using SPSS software, and statistical results are listed in table 7. As can be seen from table 7, there is a good agreement between the scores of all the paintings and the scores of the individual paintings. Thus, the average score of the two paintings was used for the next analysis.

TABLE 7 correlation statistics of painting types and rating scale scores

(2) Comfort evaluation influence law

a. Influence law of single narrowband spectral area:

the experimental data were plotted as a two-dimensional scatter plot with the area values of the narrowband spectra of different peak wavelengths as the horizontal axis and the "comfort-discomfort" (hereinafter referred to as "comfort") evaluation score as the vertical axis, for a total of 12 images, see fig. 10.

As can be seen from the graph, the area value of the narrow-band spectrum with the peak wavelength of 445nm and 525nm has great influence on comfort evaluation, and shows obvious positive correlation; the influence of the area value of the narrow-band spectrum with the peak wavelength of 595nm on the comfort evaluation has a similar rule, but a weaker positive correlation is shown; there is a weak negative correlation between the area value of the narrow band spectrum with the peak wavelength of 705nm and the comfort evaluation.

b. Law of influence of the sum of narrowband spectral areas of peak wavelengths in the same color region:

from the analysis of the influence law of the area of a single narrow-band spectrum, it can be seen that the area value of the narrow-band spectrum with certain peak wavelengths has obvious influence on the subjective evaluation result. It is thus hypothesized that energy in a region around a particular wavelength may affect the subjective evaluation result, and that this region may correspond to a particular color. According to the corresponding relation between different wavebands and colors of the visible spectrum, see table 8, the 12 kinds of narrow-band spectrums are divided according to the color areas, the area values of the narrow-band spectrums with peak wavelengths in the same color area are summed, and the influence rule of the narrow-band spectrums on subjective evaluation results is analyzed.

TABLE 8 correspondence between different bands of the visible spectrum and color

Similar to the influence rule analysis of a single narrow-band spectrum area, the sum of the values of the narrow-band spectrum areas adjacent to the peak wavelength is taken as a horizontal axis, the comfort evaluation score is taken as a vertical axis, and experimental data points are drawn in a two-dimensional coordinate system, and 6 images are taken in total, as shown in fig. 11.

As can be seen from the graph, the sum of the area values of the narrow-band spectrum in the blue region has a great influence on the comfort evaluation, and shows a clear positive correlation, and the influence of the area values of the narrow-band spectrum in the green region on the comfort evaluation has a similar rule, but has a weaker positive correlation than the influence of the blue region; the area value of the narrow-band spectrum in the red area has a weak negative correlation with the comfort evaluation; the area values of the narrowband spectrum in the purple region, the cyan region, and the orange region have little influence on the comfort evaluation score, and do not exhibit a clear correlation.

(3) Definition evaluation influence law

a. Influence law of single narrowband spectral area:

the experimental data were plotted as a two-dimensional scattergram with the area values of the narrowband spectra of different peak wavelengths as the horizontal axis and the "sharpness-blur" (hereinafter referred to as "sharpness") evaluation score as the vertical axis, for a total of 12 images, see fig. 12.

As can be seen from the graph, the area values of the narrowband spectrums with the peak wavelengths of 445nm, 525nm and 595nm have great influence on the definition evaluation, and obvious positive correlation is presented; similar rules exist for the influence of the area value of the narrow-band spectrum with the peak wavelength of 635nm on the definition evaluation, but a weaker positive correlation relationship is shown; the area values of the narrow-band spectrums with peak wavelengths of 660nm, 670nm and 705nm have a weak negative correlation with the sharpness evaluation.

and drawing experimental data points in a two-dimensional coordinate system by taking the sum of the area values of the narrow-band spectrums adjacent to the peak wavelength as the horizontal axis and taking the definition evaluation score as the vertical axis, wherein the total of 6 images are shown in fig. 13.

As can be seen from the graph, the sum of the area values of the narrow-band spectrum in the blue region and the green region has a great influence on the sharpness evaluation, and shows a clear positive correlation; similar rules exist for the impact of the area values of the narrowband spectrum in the orange region on the sharpness evaluation, but a weaker positive correlation is presented compared to the Lan Seou and green regions; the area value of the narrow-band spectrum in the red area has a weak negative correlation with the definition evaluation; the area values of the narrowband spectrum in the purple region, the cyan region, and the red region have little influence on the sharpness evaluation score, and do not exhibit a clear correlation.

(4) Vividness evaluation influence law

a. Influence law of single narrowband spectral area:

the experimental data were plotted as a two-dimensional scattergram with the area values of the narrowband spectra of different peak wavelengths as the horizontal axis and the "vivid color" evaluation score (hereinafter referred to as "vividness") as the vertical axis, for a total of 12 images, see fig. 14.

As can be seen from the graph, the area value of the narrowband spectrum with the peak wavelength of 610nm has a larger influence on the vividness evaluation, and shows an obvious positive correlation; similar rules exist for the influence of the area value of the narrow-band spectrum with the peak wavelength of 635nm on the vividness evaluation, but weak positive correlation is presented; there is a weak negative correlation between the area value of the narrowband spectrum with the peak wavelength of 445nm and the vividness evaluation.

and drawing experimental data points in a two-dimensional coordinate system by taking the sum of the area values of the narrow-band spectrums with adjacent peak wavelengths as the horizontal axis and the vividness evaluation score as the vertical axis, wherein the total of 6 images are shown in fig. 15.

As can be seen from the graph, the sum of the area values of the narrowband spectrum in the orange region has an influence on the vividness evaluation, and shows a weak positive correlation.

7. And (5) carrying out relevance analysis of influence factors.

The correlation among the area values of the narrow-band light sources with different peak wavelengths, the sum of the area values of the narrow-band light sources with the peak wavelengths in the same color region and the three-degree evaluation score is analyzed by using SPSS software, and the Pearson correlation coefficient is selected as an index. Setting conditions as factors with correlation coefficients larger than 0.5 and obvious correlation, and screening out influence factors of three-degree evaluation.

(1) Correlation analysis of comfort evaluation influencing factors

Table 9 shows the factors that have a significant effect on comfort evaluation, which are the sum of the area value of the narrowband spectrum having a peak wavelength of 445nm, the area value of the narrowband spectrum having a peak wavelength in the blue region, the area value of the narrowband spectrum having a peak wavelength in the green region (505+525+555nm), and the area value of the narrowband spectrum having a peak wavelength of 525nm, respectively. In practice, since the area value of the narrowband spectrum having a peak wavelength of 445nm is in the blue region, correlation analysis is not performed on the sum of the area values of the narrowband spectrum having a peak wavelength in the blue region; since the green region includes the area value of the narrow-band spectrum having the peak wavelength of 525nm and the correlation between the former and the comfort evaluation result is high, it is not necessary to perform correlation analysis again on the area value of the narrow-band spectrum having the peak wavelength of 525 nm.

TABLE 9 analysis of factors with significant impact on comfort evaluation

/>

* *. on the 0.01 scale (double-tailed), the correlation is remarkable

And respectively taking the area value of the narrow-band spectrum with the peak wavelength of 445nm and the area value of the narrow-band spectrum with the peak wavelength in the green area as an x-axis and a y-axis, taking the comfort evaluation score as a z-axis, and drawing a three-dimensional scatter diagram to visually represent the relevance of the influence of the two variables on the comfort evaluation, as shown in fig. 16.

As can be seen from the graph, when the area value of the narrowband spectrum with the peak wavelength of 445nm is kept unchanged, the area value of the narrowband spectrum with the peak wavelength in the green area is increased, and the change of the comfort level score is small; when the area value of the narrowband spectrum having the peak wavelength in the green region remains unchanged, the comfort score increases as the area value of the narrowband spectrum having the peak wavelength of 445nm increases. Therefore, in practical applications, an increase in comfort can be obtained by increasing the energy value in the region around 445nm in the light source spectrum.

(2) Correlation analysis of sharpness evaluation influencing factors

Table 10 shows factors that have a significant influence on the sharpness evaluation, namely, the area values of narrowband spectra having peak wavelengths of 445nm, 525nm, and 595nm, the sum of the area values of narrowband spectra having peak wavelengths in the blue region, the sum of the area values of narrowband spectra having peak wavelengths in the green region (505+525+555nm), and the sum of the area values of narrowband spectra having peak wavelengths in the orange region (595+610 nm), respectively. In fact, the blue region is the area value of the narrow-band spectrum with the peak wavelength of 445nm, so that correlation analysis is not needed for the sum of the area values of the narrow-band spectrum with the peak wavelength in the blue region; the green region contains the area value of the narrow-band spectrum with the peak wavelength of 525nm, and the correlation between the area value of the narrow-band spectrum with the peak wavelength of 525nm and the definition evaluation result is high, so that correlation analysis is not needed to be carried out on the sum of the area values of the narrow-band spectrum with the peak wavelength of 525 nm; the orange region contains the area value of the narrowband spectrum having a peak wavelength of 595nm, and the former has a low correlation with the sharpness evaluation result, so that it is not necessary to perform correlation analysis again on the sum of the area values of the narrowband spectrum having a peak wavelength in the orange region.

Table 10 analysis of factors having a significant impact on sharpness evaluation

* *. on the 0.01 scale (double-tailed), the correlation is remarkable

Three-dimensional scatter diagrams are drawn with two factors of the area value of the narrow-band spectrum with the peak wavelength of 445nm and 595nm and the sum of the area value of the narrow-band spectrum with the peak wavelength in the green area as an x-axis and a y-axis, with the sharpness evaluation score as a z-axis, and a total of 3 images are drawn to visually represent the relevance of the influence of the three variables on the comfort evaluation, as shown in fig. 17.

As can be seen from the graph (a), when the area value of the narrowband spectrum with the peak wavelength of 445nm is kept unchanged, the area value of the narrowband spectrum with the peak wavelength in the green area is increased, and the sharpness score is increased accordingly; when the area value of the narrowband spectrum with the peak wavelength in the green area remains unchanged, the area value of the narrowband spectrum with the peak wavelength of 445nm is increased, and the change of the sharpness score is small.

As can be seen from the graph (b), when the area value of the narrowband spectrum having a peak wavelength of 445nm is kept unchanged, the area value of the narrowband spectrum having a peak wavelength of 595nm is increased, and the change in sharpness score is small; when the area value of the narrowband spectrum with the peak wavelength of 595nm is kept unchanged, the area value of the narrowband spectrum with the peak wavelength of 445nm is increased, and the definition score is increased.

As can be seen from the graph (c), when the area value of the narrowband spectrum having the peak wavelength in the green region is kept unchanged, the area value of the narrowband spectrum having the peak wavelength of 595nm is increased, and the change in sharpness score is small; when the area value of the narrow-band spectrum with the peak wavelength of 595nm is kept unchanged, the area value of the narrow-band spectrum with the peak wavelength in the green area is increased, and the definition score is increased.

In summary, in practical applications, the improvement of sharpness may be obtained by increasing the energy value of the green region in the light source spectrum, or by increasing the energy value of the region near 445nm in the light source spectrum.

(3) Correlation analysis of vividness evaluation influence factors

Table 11 shows factors that have a significant influence on the vividness evaluation, which are the sum of the area values of the narrowband spectrum having the peak wavelength in the orange region and the area value of the narrowband spectrum having the peak wavelength of 610nm, respectively. In fact, the orange region includes the area value of the narrowband spectrum having a peak wavelength of 610nm, and the former has a high correlation with the vividness evaluation result, so that it is not necessary to perform correlation analysis again on the area value of the narrowband spectrum having a peak wavelength of 610 nm.

TABLE 11 analysis of factors having a significant impact on vividness evaluation

* *. on the 0.01 scale (double-tailed), the correlation is remarkable

Since only one factor affects the vividness evaluation to satisfy the screening condition, there is no need to analyze the relevance of the influencing factors. In practical applications, an increase in vividness can be obtained by increasing the energy value of the orange region in the light source spectrum.

8. Data dimension reduction processing

From the analysis results of 6 and 7, it can be seen that the area values of the narrowband spectra having peak wavelengths of 445nm, 525nm, 595nm and 610nm have a significant effect on the evaluation results of comfort, sharpness and vividness, and the sum of the area values of the narrowband spectra having peak wavelengths in the blue, green and orange areas also has a significant effect on the subjective evaluation results.

When the number of the variables exceeds three, the variables cannot be displayed by using the visual graph, so that the use simplicity of the obtained model is reduced, and the practical application is not facilitated. Therefore, it is necessary to reduce the dimension of variable data. The sum of the area values of the narrowband spectra with peak wavelengths of 595nm and 610nm is used to combine into one variable, i.e., the energy value of the peak wavelength in the orange region, according to the physical meaning of the peak wavelength.

Table 12 lists the statistical results based on analysis of the correlation between the sum of the area values of the narrowband light sources with peak wavelengths in the orange region and the subjective evaluation score using SPSS software in fig. 7. It can be seen that the peak wavelength has a significant correlation between the area value of the narrowband light source in the orange region and the sharpness, vividness evaluation score, and the pearson correlation coefficient is greater than 0.5.

Table 12 correlation statistics of area values of narrowband light sources with peak wavelengths in orange region and rating scale scores

* *. on the 0.01 scale (double-tailed), the correlation is remarkable

Finally, the number of variables was reduced to three dimensions, namely, the area value of the narrowband spectrum with a peak wavelength of 445nm, the area value of the narrowband spectrum with a peak wavelength in the green region, and the area value of the narrowband spectrum with a peak wavelength in the orange region, as input parameters for the next establishment of an evaluation model.

9. Lighting quality evaluation model

(1) Data preprocessing

a converting dependent variables into tag variables

Taking definition scoring as an example, the definition scoring mean value under each working condition is divided into the following three sections: -4 is less than or equal to the definition score mean value < -1, -1 is less than or equal to the definition score mean value < 3, and 3 is less than or equal to the definition score mean value < 4. These three intervals represent different sharpness evaluation levels, respectively "bad", "good", "excellent". The dependent variables in the three sections are labeled "1", "2" and "3", respectively, as tag variables. Comfort score, vividness score, and so on, are also classified into three classes, "poor", "good", "excellent". At this time, the model building is converted into a three-classification problem, and the distribution of the number of the data in the different definition classification labels is shown in fig. 18.

b normalizing the independent variable

The normalization of the characteristic variables is beneficial to the later data processing and improves the convergence rate in the running process of the program. Each dimension of data is mapped to a particular interval using the SVM self-contained function MinMaxScaler, typically using a default interval of 0, 1.

(2) Model building process

a dividing the data into training and testing sets.

The training set data will be used for model training and the test set data will be used for testing the accuracy of the model. Since there are 40 experimental conditions, there are 40 sets of raw data. Together, each set of data contains 3 values of the self-variable and 1 value of the factor variable. Setting a training set: test set = 4:1, the data were randomly partitioned, resulting in 32 training sets, 8 test sets.

And b, selecting parameters.

According to the target characteristics of the research, the establishment of the evaluation model belongs to the nonlinear classification problem. The kernel function was developed to solve the nonlinear classification problem and can map data from a lower dimensional space to a higher dimensional space. The most widely used gaussian kernel function (RBF) is chosen, whose kernel function is in the form of:

K(x _i ,x _j )＝exp(-γ‖x _i -x _j ‖ ² )

wherein (x_i, x_j) is training set data, x_i is a feature vector of the data, x_j is a label, and gamma is a parameter which needs to be set manually. Before model training, penalty factor C and γ values, both real numbers greater than 0, need to be specified. Thus, the parameters described above are adjusted using K-fold cross-validation and grid-based methods to obtain optimal values of C and γ.

And c, training a model.

For the data of the sharpness evaluation, the characteristic variables were 3 in total, namely, the area value of the narrowband spectrum having a peak wavelength of 445nm, the sum of the area values of the narrowband spectra having peak wavelengths of 505nm, 525nm and 555nm, and the sum of the area values of the narrowband spectra having peak wavelengths of 595nm and 610nm, respectively, and the objective variable was the sharpness evaluation level obtained by the processing. Setting parameters C and gamma as optimal values obtained in the second step, inputting characteristic variables and target variables of a training set, and training to obtain a definition evaluation model based on a support vector machine. The comfort evaluation model and the vividness evaluation are the same, wherein for the comfort evaluation data, 2 characteristic variables are the sum of the area value of a narrow-band spectrum with the peak wavelength of 445nm and the area value of a narrow-band spectrum with the peak wavelength of 505nm, 525nm and 555nm, the target variable is the comfort evaluation level obtained through treatment, 1 characteristic variable in the vividness evaluation data is the sum of the area value of a narrow-band spectrum with the peak wavelength of 595nm and 610nm, the target variable is the vividness evaluation level obtained through treatment, the obtaining mode of setting the values of parameters C and gamma is the same as that of the second step, the characteristic variable and the target variable of a training set are input, and the comfort evaluation model and the vividness evaluation model based on a support vector machine can be obtained through training.

(3) Mathematical model

And inputting the characteristic variables of the test set data into the trained model, and calculating the predicted value. The accuracy of the model can be characterized by calculating the difference between the predicted value and the target variable of the test set by the function metrics. Through calculation, the accuracy of the comfort evaluation model, the definition evaluation model and the vividness evaluation model are respectively 0.75, 0.875 and 0.875.

In addition to accuracy calculations, multi-class models are typically evaluated using a subject work characteristic curve (ROC curve). The method can intuitively display the accuracy and the sensitivity of the evaluation model when different thresholds are set. Taking the sharpness evaluation model as an example, fig. 19 plots its ROC curve. As can be seen from the figure, the model has better classification performance.

The painting illumination quality evaluation model based on the support vector machine established through the process can predict the overall feeling when the painting is watched under any exhibition Chen Guangyuan and compare the visual effect quality when different light sources irradiate the painting. The evaluation model has good operation simplicity.

FIG. 20 shows a specific application flow of the evaluation model, which requires the following three steps:

1. The spectrum to be measured is reproduced using a LEDCube. The spectral power distribution of the light source to be evaluated is input into the LEDCUP, the illuminance value of the spectral power distribution is set, the SPD mode-Shape is selected under the 'light distribution' function menu of the LEDCUP operation interface, and manual adjustment can be performed by using the 'feed back' in the process so as to obtain the spectrum with the most similar Shape.

2. And reading the proportion value of each narrow-band light source composing the spectrum at the LEDCUP operation interface, and calculating the area value of the narrow-band light source according to the function of irradiance and the proportion value. The characteristic variables input are the area value of the narrowband spectrum with the peak wavelength of 445nm, the sum of the area values of the narrowband spectrums with the peak wavelengths of 505nm, 525nm and 555nm, and the sum of the area values of the narrowband spectrums with the peak wavelengths of 595nm and 610 nm.

3. And inputting the three characteristic variables into a model, and calculating to obtain the evaluation scores of the spectrum to be measured in three dimensions of comfort, definition and vividness. According to the variable labeling process in 9- (1) -a, the evaluation grade of the spectrum to be tested in three dimensions of comfort, definition and vividness can be obtained.

The foregoing is merely illustrative of the present invention, and the present invention is not limited thereto, and any changes or substitutions easily contemplated by those skilled in the art within the scope of the present invention should be included in the scope of the present invention.

Claims

1. A painting illumination visual evaluation method related to spectrum power distribution of a light source is characterized in that, aiming at a museum painting cultural relic sample, a direct relation between spectrum power distribution SPD and subjective evaluation results is established through a visual evaluation experiment, and a spectrum specific wave band with obvious influence on illumination quality evaluation results is determined; then, through data analysis and excavation, an SPD-based museum drawing illumination quality evaluation model is established, and drawing illumination visual evaluation is realized;

the method comprises the following specific steps:

1) Using LEDCube as an illumination light source, and operating through computer software;

2) And carrying out spectrum construction and iteration by using ten narrowband spectrums to obtain a final 40-item target spectrum, wherein the peak wavelengths of the ten narrowband spectrums are respectively as follows: 447. 475, 500, 519, 555, 595, 625, 635, 658, 733; the half bandwidths are respectively: 20. 20, 30, 20, 30, 20;

screening modes meeting target spectrums are screened:

3) Acquisition of any narrowband spectral area:

4) Establishing two-dimensional graphs of areas of narrowband spectrums with different peak wavelengths and evaluation scores according to experimental data, further determining specific spectrum bands influencing the evaluation through correlation analysis, establishing two-dimensional graphs of different colors and evaluation scores, and further determining specific colors influencing the evaluation through correlation analysis;

5) Reducing the number of variables into three dimensions through correlation analysis, and realizing normalization processing of independent variables by using a function MinMaxScaler in a support vector machine; based on the conversion of the dependent variable into the label variable, the model establishment is further converted into three types of problems, and the comfort evaluation model, the definition evaluation model and the vividness evaluation model are established through data division and model training.

2. The method for visual evaluation of painting illumination related to spectral power distribution of light source according to claim 1, wherein the comfort evaluation model, the definition evaluation model and the vividness evaluation model are established through data division and model training, and the method comprises the following specific steps:

(1) Data preprocessing

a converting dependent variables into tag variables

b normalizing the independent variable

(2) Model building process

a, dividing data into a training set and a testing set;

setting a training set: test set = 4:1, randomly partitioning the data;

b parameter selection

K(x _i ,x _j )＝exp(-γ||x _i -x _j || ² )

model c training

For the data of definition evaluation, the total number of characteristic variables is 3, namely the sum of the area values of narrowband spectrums with the peak wavelength of 445nm, the area values of narrowband spectrums with the peak wavelength of 505nm, 525nm and 555nm and the sum of the area values of narrowband spectrums with the peak wavelength of 595nm and 610nm, the target variable is the definition evaluation level obtained through processing, the parameters C and gamma are the values obtained in the step b, the characteristic variable and the target variable of a training set are input, and the definition evaluation model based on a support vector machine is obtained through training; for comfort evaluation data, 2 characteristic variables are the sum of the area value of a narrow-band spectrum with the peak wavelength of 445nm and the area value of a narrow-band spectrum with the peak wavelength of 505nm, 525nm and 555nm, a target variable is the comfort evaluation level obtained through processing, 1 characteristic variable in the vividness evaluation data is the sum of the area values of the narrow-band spectrum with the peak wavelength of 595nm and 610nm, a target variable is the vividness evaluation level obtained through processing, wherein the obtaining mode of setting the values of parameters C and gamma is the same as that of step b, the characteristic variable and the target variable of a training set are input, and a comfort evaluation model and a vividness evaluation model based on a support vector machine are obtained through training;

(3) Mathematical model